Subject: Process for Oxygen Descum of PMMA
From: Rashid Zia <zia@stanford.edu>
Date: Mon, 21 Feb 2005 16:49:58 -0800
Mon, 21 Feb 2005 16:49:58 -0800
Hello Everyone,

As you may know, the development of exposed PMMA is never perfect.   In 
general, small bits of PMMA in exposed areas are left on your substrate 
following the MIBK/IPA development and IPA rinse processes.   This may not 
be of concern to most people, but these unremoved bits will translate into 
surface roughness on metallized or etched features.

This past week, I ran a variety of SEM characterization studies for 100nm 
PMMA-C involving dose levels, development times, and an oxygen plasma 
descum using drytek1 in SNF.   I found that, after appropriate calibration 
of dose, a standard 30sec development in 1:3 MIBK/IPA and a 1min rinse in 
IPA produced the best features;  by itself though, this process left 
~15-25nm surface roughness features on my sample following metallization. 
(See Figure 1, without descum step)   To decrease this roughness, I 
descummed several samples in a 4-10 sec oxygen plasma using drytek1 at 
100mT and 35-55W.   All of these samples showed reduced roughness without a 
detectable  change in feature quality, and all lifted-off 
easily.   Finally, I found that a 10sec oxygen plasma at 100mT at minimal 
power for plasma formation in drytek1 is ideal for removing final portions 
of exposed resist, and minimizing surface roughness features for 
metallization.  (See Figure 2, with 10s descum step)   (Also, I have plenty 
of additional SEMs if anyone is interested.)

It should be noted that drytek1 does not like to fire a plasma at low 
power.  The minimum power for plasma formation with 100mT of oxygen varies 
between 30W and 55W depending on the precision of the tuning and the day of 
the week, apparently.   In order to standardize the process, I suggest that 
you begin the etch process at a power level well below plasma formation 
(e.g. 15W), and then slowly increase the power until a plasma forms.  At 
that point, use your own timer to count to 10s (or your desired duration) 
after which time you can stop the process manually.   (Note for 
characterizing your own samples, I suggest you try this a bunch of times 
with nothing in the chamber, and remember to monitor and record the power 
at which the plasma is formed.)

Of course, I cannot guarantee this process will work for you, but from now 
on,a 10sec descum in oxygen plasma will follow the development process for 
me.   If try this, please let me know how it works.

Thanks,
Rashid.

ps- The attached SEM images show two different samples ( with and without a 
descum step) following metallization with 2.5nm of Cr and 55nm of Au.

["image/tiff" not shown]
["image/tiff" not shown]